The best job search hack isn't applying to more roles. It's applying to fewer, better ones with a feedback loop that actually learns.
The Summary
- Margareth Felicia Gono built a custom Claude workflow to track applications, parse job descriptions, and analyze interview sentiment, landing a senior role at DoorDash after her master's degree
- She used AI as a structured feedback system, not a résumé exaggerator, identifying patterns in rejections and hesitation points during second and third-round interviews
- Her approach: treat anxiety as unprocessed data and build tools to surface the signal in what was working versus what wasn't
The Signal
Most job seekers use AI to spam applications faster. Gono did the opposite. She built a custom workflow in Claude that functioned as a private career coach and pattern detector, analyzing what got responses and what didn't. The goal wasn't volume. It was clarity.
Her system had two parts. First, she'd feed Claude a job description and ask it to extract core skills, keywords, and themes. Then she built an automation agent, her own version of an ATS compatibility score, that compared those requirements against her actual résumé. If the match was weak, she skipped the role. If it was strong, she applied with targeted adjustments. No spray and pray. No generic cover letters that sound like everyone else's.
"Anxiety is unprocessed data. I was tired of feeling bad about my search and figured I might as well learn something instead."
The second layer was sentiment analysis on her own interview performance. Around the second or third rounds, she noticed hesitation from employers. Was it her experience? Her visa status? The way she explained her background? She couldn't tell from gut feel alone. So she started feeding Claude her interview notes and asking it to identify where momentum stalled. The AI surfaced patterns she couldn't see in the moment: certain phrasing that created doubt, topics where her answers felt thin, moments where interviewers seemed to disengage.
Key insights from her workflow:
- Job descriptions that emphasized "corporate experience" flagged higher rejection rates for her as an international student with limited US work history
- Roles that valued "applied analytics" over generic "data science" matched her master's focus and got better responses
- Interview sentiment shifted when she reframed her international background as a differentiator, not a liability
This isn't about letting AI write your résumé. Gono explicitly avoided that. She didn't want Claude to exaggerate her experience or make her sound like a ChatGPT template. She used it as a mirror, a feedback system that helped her see what was actually happening in her search versus what she thought was happening. The output wasn't better lies. It was better self-awareness.
The broader implication: job seekers are building their own agent-assisted workflows because the hiring process itself is now AI-mediated on the other side. If companies use ATS systems and sentiment tools to filter candidates, candidates can use the same tools to reverse-engineer what's getting filtered out. This is the early shape of the agent labor market. Not AI doing your job. AI helping you navigate systems designed to screen you out.
The Implication
If you're job searching, stop optimizing for volume. Build a feedback loop. Track what works. Use AI to surface patterns in your own data. The companies that will hire you are already using sentiment analysis and keyword matching. You're not gaming the system by doing the same. You're just leveling the playing field.
Gono's workflow is replicable. Take a job description. Ask Claude or another LLM to extract the core requirements. Compare them to your actual experience. If the match is weak, move on. If it's strong, adjust your materials to reflect the overlap. After interviews, ask the AI to analyze your notes for hesitation points. Iterate. The goal isn't perfection. It's incremental clarity about what's landing and what's not.
The companies building agent-first hiring tools should pay attention. Candidates are already building their own agents. The next wave of recruiting isn't about screening people out faster. It's about surfacing signal on both sides.